Learning Infinite-horizon Average-reward MDPs with Linear Function Approximation.
Chen-Yu WeiMehdi Jafarnia-JahromiHaipeng LuoRahul JainPublished in: AISTATS (2021)
Keyphrases
- reinforcement learning
- function approximation
- markov decision processes
- average reward
- optimal policy
- infinite horizon
- policy iteration
- td learning
- actor critic
- stochastic games
- long run
- partially observable
- finite horizon
- function approximators
- learning tasks
- learning algorithm
- temporal difference
- policy gradient
- markov decision process
- state space
- model free
- optimal control
- average cost
- total reward
- reinforcement learning algorithms
- finite state
- multistage
- state action
- discount factor
- partially observable markov decision processes
- action selection
- markov decision problems
- supervised learning
- dynamic programming
- multi agent
- decision making
- machine learning